Computer Supported Data-driven Decisions for Service Personalization: A Variable-Scale Clustering Method

The aim of this paper is to solve the object segmentation problem designed for service personalization in the context of individual athletic events. Focusing on certain personalized characteristics of the marathon contestants, the research puts forward a discovery method based on the variable-scale clustering (PCD-VSC). This method could be employed in order to obtain object segmentation based on scale similarity measurement. A case study is created based on a real dataset related to 59 marathon events which took place in several cities between 2017 and 2018, with a total number of 14,160 contestants. The numerical experimental results show that the PCD-VSC algorithm divides marathon runners into seventeen qualified clusters based on clear competitive and preference characteristics. Hence, this method could support the managers of marathon competitions in designing and implementing a personalized service scheme for the marathon contestants. Also, in comparison with the traditional VSC, the proposed method proves the overall accuracy and efficiency in analyzing categorical dataset with duplicate attribute values.